Hidden Sources of Care Fragmentation in Heart Failure Care Transitions: A Mixed-method Network Analysis of Clinical Notes

Abstract

Introduction:

Patients with heart failure, especially those with lower socioeconomic position, are vulnerable to adverse outcomes of care fragmentation. Care coordination mitigates care fragmentation; however, a comprehensive assessment of relationships among all relevant clinicians that enable it is lacking. This explanatory sequential mixed-methods study explores how patients’ clinician network characteristics relate to their social context and clinical outcomes following heart failure care transitions.

Methods:

We conducted a bipartite social network analysis to describe clinician networks for 1269 patients first hospitalized with heart failure in a southeastern US health system. Networks were constructed using electronic health record notes 1 year before, during, and 1 year after the index hospitalization (January 2015-February 2020). We then used stratified purposive sampling to select 11 adults with diverse socioeconomic positions but similar illness severity and comorbidity. For these patients, we conducted qualitative chart reviews of their clinical notes. Each patient’s clinician network size, density, and centrality were integrated with qualitative findings to explore clinician networks’ relationship to patients’ social context and outcomes.

Results:

Patients with higher socioeconomic positions used fewer acute care services and lived longer. Their clinicians, particularly outpatient clinicians, tended to have denser and more centrally located team networks before the index hospitalization that persisted after the index hospitalization. Their telephone notes indicated more regular and reciprocal communication patterns between patients and clinicians.

Conclusions:

Early involvement and better communication among clinician networks with greater density and centrality may help to explain better care transition outcomes observed among patients with higher socioeconomic positions.

Introduction

Care fragmentation is a persistent problem in US health care. It often occurs when patients visit clinicians across multiple sites without adequate communication to coordinate care, resulting in medical errors, poor patient outcomes, and high costs.^1,2 Despite years of research, patients with heart failure (HF), particularly those with low socioeconomic position (SEP), remain prone to care fragmentation, which is associated with increased readmission, service use, and mortality.^3–5

Care coordination, defined as the “deliberate organization of care activities and information sharing among all relevant participants to ensure safer and more effective care,”⁶ reduces care fragmentation. Transitional care interventions (e.g., telemonitoring and discharge planning) commonly aim to facilitate hospital-to-home transitions, typically focus on the triad of patient, nurse coordinator, and primary care provider, and inconsistently improve HF outcomes.⁷ A more comprehensive evaluation of multidisciplinary clinician networks across settings is needed to identify novel strategies to optimize care coordination.^7,8

Social network analysis (SNA) can advance care coordination research by revealing clinician relationship structures in health systems.⁸ While SNA and electronic health records (EHR) have been used to measure care coordination,¹ no studies have evaluated clinician networks facilitating home-to-hospital-to-home transitions. This mixed-methods study aimed to examine and describe how clinician networks of patients with HF may differ and change across settings and over time.

Methods

Design

We conducted a retrospective mixed-methods study using an explanatory sequential approach, recognizing that the meaningfulness of quantitative SNA is best understood when embedded within the context of rich narratives.⁹ Given the rich social and clinical notes recorded by clinicians in EHR, we contextualized quantitative network measures with a qualitative review of clinical notes, following quantitative, qualitative, and integration phases (Figure 1).¹⁰ The Duke University institutional review board approved this study before study initiation.

Figure 1. Phases, Procedures, Products, and Frameworks Involved for the Retrospective Sequential Explanatory Approach.

Patients with contrasting socioeconomic positions (SEP) and contrasting clinician networks (Case 1, High SEP; Case 8, Low SEP)

Setting and Sample

Quantitative analysis included a cohort of 1269 adults with first hospitalization for HF in a major university health system in the Southeastern U.S. between January 2016 and December 2018.¹² After quantitatively mapping the clinician networks, we purposively selected 11 patients for in-depth, immersive exploration of their home-to-hospital-to-home transition to understand the impact of the network characteristics on their outcomes (Figure 1).

Purposive sampling aimed to identify patients with similar risk of acute service use but ensure a proportional representation of patients across racial, financial, and environmental dimensions of SEP based on the World Health Organization’s Social Determinants of Health framework and Socio-Ecological Model.^5,13,14 To control for acute services use, we used the two most crucial factors identified in a previous quantitative study of the same cohort, the Charlson Comorbidity Index and changes in the severity of illness. We included only patients with a Charlson Comorbidity Index between 2 and 4 and those with two levels of change in the 3M severity of illness score from the cohort, which were measured based on ICD-9 and ICD-10 diagnosis codes.¹² Then, we stratified the cohort into eight possible combinations of the three SEP criteria: Black or not, Medicaid use or not, and living in highly deprived areas based on Area Deprivation Index (ADI >90).^3–5 Eleven patients were determined sufficient based on having at least one patient per subgroup and information power.¹⁵ We reviewed notes on social history to verify the accuracy of these social factors.

Being Black, Medicaid users, and living in highly deprived areas were considered adverse social strata. Five patients (Cases 1–5) had 0–1 adverse social strata and were categorized as Higher SEP, while six (Cases 6–11) had multiple adverse social strata and were considered Lower SEP. Gender and marital status⁵ were not used as SEP factors to order the cases from the highest to the lowest SEP.

Quantitative Phase

EHR Data Extraction.

Structured note data (i.e., note type, time, author, and author’s clinical role, credential, and department affiliation) 1-year before, at, and 1-year after the index hospitalization were extracted from Epic Clarity by data analysts at the Heart Center. All note types (e.g., progress, telephone) created by all clinicians from all settings for the identified cohort were included. Following common measures used to assess HF transitional care quality, we measured hospitalizations, unplanned 30-day readmissions, and mortality.^2,3,7,16 Encounter data were used to count hospitalizations, including inpatient and emergency department (ED) stays.¹² Inpatient encounters within 30-day post-index hospitalizations without elective surgeries were identified as unplanned 30-day readmissions, which were validated via reviewing admission notes. We conducted data cleaning and analysis in R (Version 3.6.3).¹⁷

System and Team Network Construction.

Using bipartite SNA, based on how all clinicians write notes for shared patients, we mapped the system-level clinician networks for the whole cohort.¹⁸ A clinician was represented by a point (node), and their connections (ties) were inferred and weighted based on the minimum frequency of notes (tie weights/strength) written for shared patients.^18–20 The more they wrote notes for the same patients, the stronger their relationships. Connecting all nodes and ties using igragh (software package for SNA) in R, we constructed the system-level networks of outpatient clinician networks 1-year before and 1-year after the index hospitalization and the inpatient clinician network at the index hospitalization. We then identified the 11 patients’ clinician networks by tracking and mapping the clinicians who wrote notes for each patient and their connections from the system network.

Network Measures.

To describe clinician networks, we assessed three key measures: size (number of clinicians involved in a patient’s care), density (how closely connected clinicians are to each other), and centrality (the importance or influence of individual clinicians within the network).^18–20 Centrality was measured using adjusted valued degree centrality (indicating local influence) and Eigenvector centrality (indicating system-level influence).^18–20 For each patient, we averaged centrality scores across their care team to estimate how centrally positioned the team was within the whole system.

Qualitative Phase

We conducted a deductive content analysis of clinical notes using an a-priori codebook to describe the patients’ context for care and understand clinician roles and the formation and disruption of clinician networks. Guided by the AHRQ Care Coordination Measurement framework and Network Model for Transitional Care from a prior integrative review of transitional care interventions, we used structured codes to capture relevant information.^1,7 Steps included: (1) reviewing all encounters and notes to understand the patient’s care journey; (2) reading narrative notes for context; (3) coding social context and interactions involved for care access, management, or transitions; (4) identifying key persons in care coordination.

The notes were manually coded into matrices and independently reviewed by SW and BG to describe each case’s: (1) social factors (e.g., education, transportation concerns), (2) clinical context (e.g., complaints, reasons for visits, care plans, patient instructions), (3) family or caregiver involvement, (4) external clinicians (e.g., home health care) who did not chart in the study EHR system, (5) referrals and new clinicians engagement, (6) key clinician(s) the patient frequently visited or communicated with who have ongoing awareness of patient’s evolving context and care plan, (7) common network structures shown in transitional care interventions, such as three key actors (triad) and (8) changing or consistent communication or visit patterns with key clinicians across multiple encounters. Weekly briefing meetings were held to refine code definitions and ensure consistent application across cases.

Quantitative and Qualitative Results Integration

Network measures and graphs from quantitative analysis were integrated with qualitative codes by constructing joint display matrices. These matrices allowed comparison of participant relationships and roles with network graphs and measures to explore the meaning of quantitative network indices, and to identify important events or relationships patterned with changes in network structure. We conducted a two-level analysis: first, a within-case chronological analysis of each patient’s care events and activities and clinician involvement, and second, a cross-case comparison to identify recurring network themes related to patients’ social context and outcomes to explore how clinician networks relate to context and outcomes.

Results

Characteristics of Patients and Their Clinicians’ Networks

The 11 patients varied considerably in social characteristics, as shown in Table 1. Five patients [45%] had Medicaid; five were African American and represented 60% of the Medicaid recipients; and eight lived in highly deprived areas or were homeless. Most were female (n = 7) and unmarried (n = 8). The mean age of the 11 patients was 64.2 (SD = 15.2). Four had systolic HF, four had diastolic HF, and the others had systolic and diastolic HF or unspecified HF. For these 11 patients, 164 clinicians were involved in the index inpatient stay; between 63 and 214 outpatient clinicians were involved during the 1-year before and 1-year after the index hospitalization.

Table 1.

Patients’ Characteristics Ordered from the Highest to Lowest Socioeconomic Position

	Sociodemographic			Social Strata			Medical Strata			Utilization			Outcomes
Pt.	Age	Gender	Marital Status	ADI	Racea	Medicaid	CCI	SOI change	SOIb	# of Outpatientc	# of EDc	# of Inpatientc	30-day Readmit	Days to Deathe
1	80	Male	Married	8	White	No	3	2	Major	47	0	2	No	Alive
2	47	Male	Married	95	White	No	2	2	Major	17	0	7	Planned	Alive
3	>85	Female	Widowed	23	White	No	2	2	Extreme	14	0	2	No	421
4	69	Female	Widowed	91	White	No	3	2	Moderate	20	0	3	Planned	Alive
5	61	Female	Single	33	Black	No	4	2	Major	7	0	5	No	Alive
6	66	Male	Divorced	90	Black	No	3	2	Extreme	6	11	2	Unplanned	Alive
7	57	Male	Widowed	NAd	White	Yes	2	2	Moderate	0	8	9	Unplanned	183
8	61	Female	Single	NAd	White	Yes	4	2	Major	38	3	9	No	310
9	46	Female	Married	90	Black	Yes	3	2	Moderate	48	7	5	Unplanned	681
10	61	Male	Single	NAd	Black	Yes	2	2	Major	0	2	2	No	Alive
11	58	Female	Single	98	Black	Yes	3	2	Major	13	13	10	Unplanned	206

Note: ADI = Area Deprivation Index, CCI = Charlson Comorbidity Index, ED = Emergency Department, HF = heart failure, NA = Not applicable, Pt. = Patient, SOI = Severity of Illness

^aAll participants were identified as non-Hispanic.

^bSeverity of illness 3M score at the time of the index hospitalization

^cTotal number of outpatient, ED stay, and inpatient stay visits during the study period.

^dCases 7 and 10 were homeless, and Case 8 lived in a trailer home in a trailer park based on EHR social notes. Due to unstable living situations, their ADI may not be accurate but was classified as high based on their social context.

^eNumber of days from discharge date of the index hospitalization to death date. “Alive” indicates mortality status as of December 2021.

Patients with higher SEP had more densely connected and centrally positioned clinician networks in inpatient and outpatient settings (Table 2 and 3). These network characteristics co-occurred with reduced hospitalizations and lower mortality 1-year after index hospitalization in patients with higher SEP compared to those with lower SEP. Differences in clinician network characteristics across patients with diverse SEP were evident before the index hospitalization and persisted thereafter (Tables 2 and 3).

Table 2.

Characteristics of Patient’s Clinician Network Structure by Patient Case

	Outpatient Clinician Network 1 Year before Index		At Index Hospitalization		Outpatient Clinician Network 1 Year after Index
Patient	Size	Density	Size	Density	Size	Density
1	14	0.76	9	0.94	24	0.83
2	12	0.44	33	0.73	49	0.79
3	11	0.86	18	0.39	16	0.48
4	6	0.20	17	0.64	28	0.61
5	8	0.25	9	0.36	11	0.55
6	0	N/A	3	0.67	14	0.56
7	0	N/A	31	0.32	10	0.47
8	13	0.55	8	0.36	64	0.33
9	0	N/A	7	0.62	35	0.65
10	0	N/A	14	0.07	1	0
11	19	0.25	15	0.05	33	0.21

Table 3.

Characteristics of Each Patient’s Clinicians’ Network Position by Patient Case

	Outpatient Clinician Network 1-year before Index (Mean ±SD)		At Index Hospitalization (Mean ±SD)		Outpatient Clinician Network 1-year after Index (Mean ±SD)
Patient	Adjusted Valued Degree Centrality	Eigenvector Centrality	Adjusted Valued Degree Centrality	Eigenvector Centrality	Adjusted Valued Degree Centrality	Eigenvector Centrality
1	91.76 ±45.62	0.14 ±0.08	107.63 ±47.86	0.38 ± 0.34	131.99 ± 51.64	0.13 ±0.19
2	61.25 ±119.31	0.08 ±0.10	65.26 ±34.87	0.14 ± 0.12	195.12 ±98.39	0.25 ±0.25
3	47.62 ±25.73	0.003 ±0.003	25.40 ±16.56	>0.001 ± >0.001	81.87 ±43.20	0.01 ±0.04
4	62.03 ±55.21	0.009 ±0.007	58.20 ±28.31	0.16 ±0.21	117.11 ±84.23	0.16 ±0.25
5	49.43 ±45.27	0.14 ±0.28	19.95 ±9.17	>0.001 ± >0.001	123.95 ±79.08	0.16 ±0.29
6	N/A	N/A	18.16 ±11.08	>0.001 ± >0.001	91.82 ±67.67	0.10 ±0.24
7	N/A	N/A	16.71 ±12.76	>0.001 ± >0.001	154.42 ±122.97	0.05 ±0.08
8	48.18 ±30.39	0.002 ±0.003	18.83 ±14.76	>0.001 ± >0.001	101.45 ±71.93	0.03 ±0.07
9	N/A	N/A	26.40 ±12.00	>0.001 ± 0.001	129.48 ±84.67	0.10 ±0.18
10	N/A	N/A	24.79 ±41.20	0.01 ± 0.03	81.23 ±NA	0.18 ±NA
11	32.80 ±34.41	0.02±0.04	6.58 ±10.34	0.001 ± 0.001	82.39 ±79.44	0.02 ±0.05

Patients with lower SEP had higher inpatient and ED use, started outpatient use later, and died sooner (Table 1). Cases 6–11 accounted for most ED visits (2–13 times), and several had unplanned 30-day readmissions (Cases 6, 7, 9, and 11) or died within two years post-discharge (Cases 3, 7, 8, 9, and 11).

Patients’ Clinician Network Density.

Patients with higher SEP tended to have clinician networks that were more densely connected before the initial HF hospitalization and afterward (Table 2). The density of clinician networks for patients with lower SEP was much more likely to be zero (Cases 6,7,9,10) or low (Case 11) before the initial HF hospitalization. All patients who had a decrease in outpatient clinician network density (Cases 3, 8, 11) died within two years.

Patients’ Clinician Network Size.

Network sizes varied greatly and did not show an obvious relationship with outcomes. Outpatient network size tended to be zero among patients with lower SEP (4 out of 6) before the index hospitalization (Table 2) and tended to increase after the index hospitalization. The outpatient clinician network size was smaller than the inpatient network size. The distribution of inpatient network size for the 11 cases ranged from 3 to 153, 3 to 33, and 7 to 266, before, during, and after the index hospitalization, while the outpatient network size ranged from 0 to 19 before and 1 to 64 after the index hospitalization.

Clinicians’ Centrality.

Patients with higher SEP tended to have clinician teams composed of clinicians more centrally positioned within the health system before the index hospitalization (Table 3). The average centrality of each patient’s clinician team generally increased after the index hospitalization, particularly for adjusted valued degree centrality. Centrality values were best understood when assessed in the context of their relative magnitudes across cases. Patients with higher centralities, especially Eigenvector centrality, for both inpatient and outpatient clinicians, experienced fewer hospitalizations and longer survival.

How Network Characteristics Relate to Context and Outcome

Care Coordination Triad.

When the patient and caregiver[s], the PCP team, and the cardiovascular specialty team engaged in consistent and reciprocal communication, better outcomes followed. Cases 1 and 2 exemplified the strength of this triad. EHR notes indicated spousal support in care and clinician networks characterized by consistent and reciprocal communication across teams. For example, Case 1 had consistently high outpatient clinician network density before (0.76) and after (0.83) the index hospitalization. Case 2 established similar stable and reciprocal communication patterns after the index hospitalization, which corresponded to improved and high network density (0.79) after the index hospitalization. Their clinician networks had the highest density during the index hospitalization and were more densely connected with outpatient clinicians before and after the index hospitalization than others (Table 2). Telephone and MyChart message notes showed continuous conversations for medication refills, questions, self-monitoring, and symptom management. Cases 1 and 2 had PCPs outside the health system, but the clinical notes mentioned communications between the PCP and cardiology teams. Although Case 1 was among the oldest patients in the sample, he had the fewest number of hospitalizations, no unplanned hospitalization, and remained alive.

However, when the patient and caregiver[s], the PCP team, and the cardiovascular specialty team were not actively engaged, care management and outcomes varied. For Cases 6–11, who had intersections of 2 or more negative social strata, having all three roles present in care was rare. Most had no PCP or cardiologist before the index hospitalization (Cases 6, 7, 9, 10). Most lacked caregiver support (Cases 6, 7, 8, 10), and some had had unstable living situations, such as being homeless (Cases 7 and 10), living in a trailer home (Case 8), living in a sibling’s home (Case 11), or being incarcerated (Case 10). Limited family or social support was noted to create barriers to care, such as difficulty attending appointments or pharmacies due to lack of transportation (Cases 6, 7, 8). Case 7 told a clinician that it was hard for him to follow dietary recommendations because “[I] can’t get to the grocery and can’t walk far, no one to help,” and another patient (Case 9) reported “being homeless at this time, making diet adherence difficult. His diet consists of fast-food meals.” Although Cases 3, 4, 6, 9, 11 exhibited all three components of the triad, evidence of consistent reciprocal communication with clinicians was absent.

Case Comparison of Network Density and Centrality.

To investigate what might explain the difference in network characteristics aside from the lack of outpatient care teams, we compared four cases (Cases 1, 4, 8, 11) at polar ends of the spectrum of SEP but who had similar types of clinicians involved in their care throughout the study period (i.e., consistent PCP teams and regularly accessed outpatient specialty services including cardiovascular specialists). Care outcomes differed markedly. Cases 1 and 4 had no ED visits and few (mostly planned) hospitalizations and survived until the end of the study period. In contrast, Cases 8 and 11 had frequent ED and inpatient stays, and Case 11 had unplanned 30-day readmission after the index hospitalization. Both patients died within a year of the index hospitalization. Compared to Cases 8 and 11, Cases 1 and 4 had stronger clinician networks (higher network density), and more inpatient and outpatient clinicians in central positions in the clinician system network.

To demonstrate differences in clinician density and eigenvector centrality, we visualized the clinician team networks (Figure 1a) for Cases 1 and 8 at their index hospitalization and traced where the clinicians were in the system network (Figure 1b). Cases 1 and 8 had similar clinical needs and similar numbers and types of clinicians at the first HF hospitalizations, but Case 8 was cared for by clinicians in less central positions and in weakly connected networks, which persisted after the index hospitalization. Case 1’s clinicians also had stronger connections measured by tie weights (shown as tie thickness) and much higher eigenvector centrality (node size in Figure 1). Although Case 1 was 19 years older than 8, Case 1 remained alive, while Case 8 died 310 days post-discharge, with 5 times the hospitalizations of Case 1.

Longitudinal Network Changes.

Longitudinal comparison showed two common scenarios: patients who successfully found key clinician(s) who guided their care management and connected them with essential resources, and patients who struggled to find consistent advocates in their clinician network. These two scenarios consistently corresponded to differences in outcome, with weak clinician networks (i.e., weakly connected structure, peripheral system position, or deteriorating network) more prevalent among patients in lower SEPs with worse care coordination, more unplanned hospitalizations, and shorter survival.

Cases 3 and 4 both had regular family, primary care, and specialist support; their network changes highlight how networks impact care. Case 3’s network density decreased after losing her primary care physician, which was followed by medication error, weak communication, multiple HF hospitalizations, and death in the hospital 1-year post-index hospitalization. Case 4’s network density and centrality increased post-hospitalization, which co-occurred with communication and care management pattern changes. Despite post-surgery complications, Case 4 shifted from minimal communication with her cardiologist to consistent and reciprocal communications with a new cardiologist-surgeon team, leading to a structured and continuous outpatient care plan, and she remained alive.

Patients with lower SEP struggled to establish or maintain stable relationships with outpatient care teams, showing consistently low network density and Eigenvector centrality. Even with family care support, consistent PCPs, and access to services like home health care and specialists, these patients often relied heavily on acute care services. For example, Case 11 had multiple ED visits and inpatient stays due to symptom burdens despite an increase in her outpatient clinician network size (from 19 to 33) and diversity. Her network density remained consistently low (0.25 and 0.21) before and after the index hospitalization. Her clinicians were consistently positioned at the periphery of the system.

Discussion

This study used a novel approach to combine EHR metadata with clinical notes to describe individual patients’ clinician networks across settings and over time for patients with HF. These results also inform the generation of hypotheses of the relationships between clinician networks, patients’ SEP and HF transitional care outcomes, that are amenable to testing in future studies building on our findings. Patients with higher SEP tended to use fewer acute care services, have more planned inpatient stays, and live longer. Their respective clinician teams, particularly outpatient clinicians, tended to have higher network density and be more centrally located in the health system network, and they exhibited reciprocal communication patterns with outpatient clinicians.

These results support the existence of a complex interplay between individual and interpersonal factors influencing health outcomes, as theorized in the Socio-Ecological Model.¹⁴ Previous research on interpersonal factors has primarily focused on informal care networks, such as caregiver support.²¹ People with low SEP tend to have smaller networks, weaker ties, and less frequent contact with their informal care network, which is associated with worse outcomes.^21,22 This study adds evidence that clinician relationship structure (formal care network) interacts with individual context and also influences health outcomes. An interplay may exist between informal and formal care networks; for example, limited informal support among patients with low SEP often hinders follow-up appointments or pharmacy access.

How care is configured may be as important as the frequency of care interactions and access.²³ Our study adds to McWilliams’ early reports showing that patients with low SEP delay accessing outpatient services and, even when care was accessed comparable to those with high SEP, patients with low SEP tended to interact with weaker and more peripheral clinician networks. In contrast, patients with high SEP were cared for by well-connected clinicians. These clinicians may be more likely to use social capital (i.e., social relationships that can provide resources to help achieve desired outcomes) to accelerate patients’ access to better quality care,²¹ and positively influence system function, connectivity, and information consistency.^20,24 While studies have shown that clinician network characteristics influence clinician performance, few have linked these characteristics with patients’ social context and health outcomes.^19,25

Lastly, our findings advance empirical evidence supporting reciprocal communication patterns among patients and families, PCP teams, and cardiology teams for effective care coordination. Standard transitional care processes, such as comprehensive discharge planning, follow-up calls, and scheduling of PCP and cardiologist follow-ups, rarely foster consistent and reciprocal relationships with ambulatory teams. This gap is associated with poor management and patient uncertainty about whom to contact during deterioration.²⁶ A reciprocal or bidirectional network communication was characterized by high network density and continuous and regular telephone or MyChart notes with primary and specialty care teams over time, which indicates strong relationships among the triad.^20,27 These findings underscore how communication patterns and clinician relationships, particularly in outpatient settings, influence care outcomes—a growing focus in the pay-for-performance era. Likewise, interventions to improve relationships among patients and caregivers, specialists, and primary care teams may reduce care fragmentation and costs,^7,28 such as communication interventions at discharge or relational coordination interventions that have shown effectiveness.^29,30

Limitations

This study was limited to clinicians within one health system due to EHR interoperability issues. External clinicians were not mapped, but we reviewed notes and fax records to assess their involvement and for cross-case comparisons. Findings from a southern U.S. health system may not generalize elsewhere. Our sample included only patients with HF; our patients’ clinician networks may have underestimated the role of non-cardiovascular clinicians, especially PCPs. We did not explore differences between public and private insurance beyond Medicaid, which may have missed an opportunity to understand differences in clinician networks among public vs. private insured patients. While we followed best SNA practices, relationships inferred from shared notes may not fully capture informal interactions, such as hallway discussions.

Conclusions

This study demonstrated the feasibility of constructing clinician networks using HER notes to better understand relational processes in HF care transitions. Using SNA reveals the historically invisible web connecting patients and clinicians, representing a new direction to advance care coordination research. Findings suggest that the structure and interactivity of clinician networks across care settings may improve care outcomes. A larger, hypothesis-testing study is warranted to evaluate how clinician network characteristics influence health outcomes.